As artificial intelligence (AI) models become more advanced, YouTube is stepping up its game by automatically labeling videos that use significant photorealistic AI. The change comes as the platform aims to take a more active role in policing content on its site.

The move follows Google’s release of Gemini Omni, a new family of multimodal AI models capable of creating high-quality videos with a deep understanding of physics and culture. Starting in May, YouTube will use internal signals to identify AI-generated content and label it accordingly, although creators are still encouraged to disclose their use of AI.

Labels will now be more visible, appearing directly below the video player for long-form videos and overlaying Shorts. For slightly altered or unrealistic content, such as a prancing unicorn, labels will remain in the expanded description only. Notably, these labels won’t affect how a video is recommended or its ability to monetize.

YouTube’s policy has been updated to include stricter enforcement of AI labeling rules, particularly for content that might be mistaken for real people or events. While creators can update disclosure status if their content was misidentified, they cannot remove the labels if created using YouTube’s own AI tools like Veo or Dream Screen.

The company is also investing in AI for other features, including interactive search and playlist generation, while expanding its deepfake detection capabilities to allow anyone to scan for face matches on the platform. This latest development suggests a growing emphasis on transparency and responsibility as AI technology continues to evolve.